Modeled Behavior has added some analysis about the housing boom. Here's their graph showing the number of new housing units per the number of additional population:
Considering the shrinking size of the average household, I'm surprised to see how low this ratio has been since 1990. And, it's surprising to see how low it was, even at the peak in 2005.
So, even if my earlier analysis is correct, that prices in the 2000's boom were reasonable, why did the new demand for housing lead to higher prices instead of increased quantity? Many areas of the depopulated inner parts of the country didn't see the price gains of the coasts and the growing metropolises, so maybe we have reached an era of development where location is important enough that we are willing to bid up the price of prime land. Or, maybe strict zoning regulations are the issue, restricting the availability of new homes.
But, Las Vegas and Phoenix were epicenters of the housing boom, and these are both areas with relatively lax zoning restrictions, and significant growth potential, where much of the new building was suburban and exurban. So, both of those explanations seem lacking.
At the height of the boom, demand for new tract homes in the Phoenix area was greater than the rate at which builders were able to permit new lots, so for a while there were weekly lotteries to determine which buyers could purchase homes in some developments. I don't know if the bottleneck was regulatory or organizational, but there was a point when the demand for new homes was greater than the ability to supply them, even in the metro area that had managed decades of exponential growth.
This leaves the mystery of why this was the case when new housing units on a national scale were not at a particularly high point, compared to earlier periods.
There are a lot of seeming incongruities here. I haven't found the narrative that ties them all up into a nice bow yet.